Web-Scale WorkflowEditor: Schahram Dustdar dustdar@dsg.tuwien.ac.at

july/AuGuST 2014 1089-7801/14/$31.00 2014 IEEE Published by the IEEE Computer Society 55

In the past few years, cloud computing has gained considerable momentum as a new computing paradigm for provisioning diverse services. With cloud computing, large-scale, dis-tributed workflow applications can aggregate services and scalable computing resources on demand with practically no capital investment and modest operating costs.15 Despite a consid-erable amount of research on addressing various cloud computing challenges, cloud services dis-covery remains an untouched area.2,4

Indeed, in the context of cloud computing, we must revisit service discovery challenges for several reasons (see the sidebar for more research on this topic). First, cloud services are offered at different levels. Currently, at least three different service levels are available software as a service (SaaS), platform as a service (PaaS), and infra-structure as a service (IaaS). Second, the lack of standards for describing and publishing cloud ser-vices makes discovering them even harder. Unlike Web services, which use standard languages such as the Web Services Description Language (WSDL) to expose their interfaces and UDDI to publish their services to registries, the majority of publicly available cloud services arent based on descrip-tion standards,2 making cloud service discovery

problematic. For example, some publicly avail-able cloud services (such as Dropbox) dont men-tion cloud at all, whereas some businesses that have nothing to do with cloud computing (such as cloud9carwash; www.cloud9carwash.com) might use cloud in their names or service descriptions.

Several interesting questions center on cloud services discovery:

How do we identify whether a service on the Web is a cloud service?

How many cloud services are currently avail-able on the Web, and who provides them (that is, are cloud services provided only by major vendors such as Microsoft, IBM, Amazon, Google, and so on)?

What kind of cloud service providers are on the Web?

From which part of the world are cloud ser-vices provisioned?

To what extent do established service-ori-ented computing (SOC) standards contribute to cloud computing?

To what extent do consumers trust cloud services?

Is there any publicly available cloud service dataset for use in cloud computing research?

Analysis of Web-Scale Cloud ServicesTalal H. Noor Taibah University

Quan Z. Sheng University of Adelaide

Anne H.H. Ngu Texas State University

Schahram Dustdar Vienna University of Technology

Cloud services have unique characteristics, including dynamic and diverse

service offerings at different levels, few standardized description languages,

and varied deployment platforms. Searching such services is thus challenging.

The authors cloud service crawler engine collects metadata about 5,883 cloud

services over the Web after parsing more than half a million possible links. An

extensive statistical analysis on this data gives an overall view of cloud service

provisionings current status.

IC-18-04-WSWF.indd 55 30/05/14 5:30 PM

Web-Scale Workflow

56 www.computer.org/internet/ IEEE INTERNET COMPuTING

Here, we describe our design of a cloud services crawler engine (CSCE)and report our statistical analysis on 5,883 real cloud services collected from the Web.

Cloud Service CrawlerEngineOur CSCE crawls search engines and collects cloud service information avail-able on the Web. Figure 1a shows the CSCEs system architecture, which con-sists of six layers.

The cloud service providers layer (top right in Figure 1a) consists of

different cloud service providers who publicly provision and advertise their services on the Web. These cloud ser-vices are accessible through Web por-tals and indexed on search engines such as Google, Yahoo, and Baidu. Some websites, such as Cloud Hosting Reviews (http://cloudhostingreview.com.au) and Cloud Storage Service Reviews (http://online-storage-service-review.toptenreviews.com) let users provide feedback. The potential set of cloud service providers that the vari-ous search engines index form the initial input to the crawler.

The cloud services ontology layer maintains the cloud services ontology (CSO), which contains a set of con-cepts and relationships that let the crawler automatically discover, vali-date, and categorize cloud services. This layer maintains the ontology via the ontology updater module.

The cloud services seeds collection layer collects possible cloud service seeds (that is, their URLs). The seed collector module considers several possible resources in search engines, such as indexed webpages, WSDL and Web Application Description

Related Work in Cloud Services Discovery

Service discovery is a fundamental approach in several research areas, including ubiquitous computing, mobile ad hoc networks, peer-to-peer (P2P), and service-oriented computing.13 However, with the advent of the cloud, we must reconsider challenges in this area because solutions for effective cloud service discovery are limited.1,4

Some researchers propose ontology techniques for cloud services discovery. One study proposes a cloud service discovery system (CSDS) that exploits ontology techniques to find cloud services that are closer to consumers requirements.4 Here, agents perform reasoning methods such as similarity, equivalent, and numerical reasoning. unfortunately, this work is only vali-dated in a small, simulated environment. We conducted our work across the entire Web to discover real cloud services. In addi-tion, our cloud services ontology design follows the uS National Institute of Standards and Technology (NIST) cloud computing standard, which helps in filtering out noisy data and increasing discovery results accuracy.

Other researchers propose using distributed hash tables (DHTs) for better discovery and load balancing of cloud services. One study presents the concept of a cloud peer that extends DHT overlay to support indexing and matching of multidimen-sional range queries for service discovery.5 This approach is validated on a public cloud computing platform (Amazon EC2). The authors work focuses on a closed environment. In contrast, we focus on discovering cloud services on an open environment (that is, the Web) to let any users or applications search cloud services that suit their needs.

Discovering Web services has been an active research area with some good results. One work collects Web Services Descrip-tion language (WSDl) documents by crawling uDDI business registries (uBRs) as well as search engines such as Google, yahoo, and Baidu.2 The authors present some detailed statistical informa-tion on Web services, such as active versus inactive Web services

and object size distribution. Another study collects Web services data through Google API and presents some interesting statisti-cal information related to Web services operation, size, word distribution, and function diversity.6 Most recent are findings on the current status of RESTful Web services.7 The authors use 17 different RESTful service design criteria (for example, availability of formal description) to analyze the top 20 RESTful services listed on the ProgrammableWeb (www.programmableweb.com). unlike previous work that discovers Web services by simply col-lecting interface documents (such as WSDl files) and searching uBRs, discovering cloud services presents more challenges, such as the lack of standardized description languages for cloud ser-vices, which need full consideration.

References1. y. Wei and M.B. Blake, Service-Oriented Computing and Cloud Comput-

ing: Challenges and Opportunities, IEEE Internet Computing, vol. 14, no. 6,

2010, pp. 7275.

2. E. Al-Masri and Q. Mahmoud, Investigating Web Services on the World

Wide Web, Proc. 17th Intl Conf. World Wide Web, 2008, pp. 795804.

3. E. Meshkova et al., A Survey on Resource Discovery Mechanisms, Peer-

to-Peer, and Service Discovery Frameworks, Computer Networks, vol. 52,

no. 11, 2008, pp. 20972128.

4. j. Kang and K.M. Sim, Towards Agents and Ontology for Cloud Service

Discovery, Proc. 2011 Intl Conf. Cyber-Enabled Distributed Computing and

Knowledge Discovery, 2011, pp. 483490.

5. R. Ranjan et al., Peer-to-Peer Cloud Provisioning: Service Discovery

and load-Balancing, Cloud Computing: Principles, Systems, and Applications,

Springer, 2010, pp. 195217.

6. y. li et al., An Exploratory Study of Web Services on the Internet, Proc.

IEEE Intl Conf. Web Services, 2007, pp. 380387.

7. D. Renzel et al., Todays Top RESTful Services and Why They Are

Not RESTful, Proc. Web Information Systems Eng., lNCS 7651, 2012,

pp.354367.

IC-18-04-WSWF.indd 56 30/05/14 5:30 PM

Analysis of Web-Scale Cloud Services

july/AuGuST 2014 57

Language documents (WADL is the REST equivalent of WSDL used to describe RESTful Web services), and advertisements. To collect data, the seed collector uses some of the con-cepts in the first few levels of the cloud services ontology as keywords (cloud services, IaaS, PaaS, SaaS, and so on), then sends the collected seeds to the cloud services filtration layer for validation.

The cloud services filtration layer filters the seeds collected from the seed collector. The cloud services veri-fier first determines whether a cloud services seed is active or inactive. Inactive seeds are kept in the inactive cloud services database for further

checking (some inactive seeds might just be temporarily unavailable), and the error codes are also captured. Active seeds are passed to the cloud ser-vices validator, which validates them using concepts from the cloud ser-vices ontology. For example, if the seeds webpage contains concepts that are related to cloud services, such as IaaS, storage, and infrastructure, then the seed is considered valid. However, if the seeds webpage contains other concepts, such as news, article, paper, or weather, then the seed is invalid because the collected seed could be a news website that publishes articles about cloud services. Invalid seeds are kept in the invalid cloud service

database, whereas valid seeds are cat-egorized (into IaaS, PaaS, or SaaS) before passing to the next layer.

The cloud services data extrac-tion layer extracts information for active and valid cloud services (for example, cloud service ID, URL, and description). The data is stored in the corresponding databases in the cloud services storage layer for further sta-tistical analysis.

Cloud Services OntologyThe CSO provides the crawler engine with meta-information and describes cloud services common data seman-tics, which is critical in the sense that cloud services might not necessarily

WWW Service provisioningand advertisements

Cloud serviceproviders

Cloud servicecrawler engine

Cloud servicesseed collection

Seedcollector

Cloud servicesextractor

Cloud servicesltration

Cloud servicesverier

Cloud servicesvalidator

Cloud servicesontology

Ontologyupdater

Cloud servicesdata extraction

Cloud servicesstorage

IaaS

PaaS

SaaS

NetworkBandwidth

OnlineBackup VMMachineImage

ApplicationServer

FileSystem

VirtualMachine

OS

Infrastructure

WebHosting

PaaSSaaS

Cloudservices

SocialNetworking

ReviewPost

NewsResearch

WikiConsulting

Training ArticlePaper

Climate

AnalysisWeather Feedback

Blog

FaceBook

YouTube

Qualitative

Quantitative

Statistics

Conference

is-a

is-not-a

Root node

Concept

Relations Nodes

ReportMagazine

Journal

IaaS

Communication Storage

DatacenterNetworkLatency

Seed collectionand data extraction

Active andvalid cloud

service seeds

Inactivecloud

services

Invalidcloud

servicesConcepts

and relationsTrust feedback

Cloud serviceswith advertisements

WSDL- and WADL-based cloud services

Cloud services

Cloud services Web portalsSearch engine advertisementsCloud service review websites

(a) (b)

Figure 1. Cloud service crawler engine (CSCE). (a) The system architecture consists of six layers. (b) The cloud services ontology (shown in part) lets the CSCE collect possible cloud service seeds and filter out invalid ones.

IC-18-04-WSWF.indd 57 30/05/14 5:30 PM

Web-Scale Workflow

58 www.computer.org/internet/ IEEE INTERNET COMPuTING

use identity words (cloud, infrastruc-ture, platform, software, and so on) in their names and descriptions. When developing the CSO, we considered the common concepts that appear in the cloud computing standard from the US National Institute of Standards and Technology (NIST; http://csrc.nist.gov/publications/drafts/800-146/Draft-NIST-SP800-146.pdf).

Our CSO contains a set of concepts and relationships between concepts that lets the CSCE automatically dis-cover, validate, and categorize cloud services on the Web. We developed the CSO based on the Protg Ontol-ogy Editor and Knowledge Acquisi-tion System (http://protege.stanford.

edu), which we used to construct the ontology and reason over the con-cepts. These concepts let the CSCE collect possible cloud service seeds and filter out invalid ones. The CSO defines two different relations: is-a and is-not-a. For instance, the seed collec-tor uses the concepts that are associ-ated with is-a relations (the top part of Figure 1b) to collect possible cloud service seeds from search engines. On the other hand, the cloud services vali-dator uses concepts that are associated with is-not-a relations (the bottom part of Figure 1b) for cloud services valida-tion. Finally, the cloud services valida-tor uses the concepts that are associated with is-a relations to categorize a valid

cloud service as IaaS, PaaS, SaaS, or a combination of these models.

Statistical Analysis and ResultsWe present a comprehensive, statisti-cal analysis of the collected data on cloud services, from several different aspects. These results also provide some insight into the questions we presented in the introduction.

Cloud Services IdentificationTo optimize the crawling performance, we used three different instances of the CSCE (each instance collects the data using multiple threads) to run simultaneously from three different machines. At an early stage, we con-figured the crawler to crawl up to five levels deep in a potential cloud services website. However, we discontinued this because its time consuming, and no significant difference exists in the crawling results. Therefore, we config-ured the crawler to crawl the first level of a potential cloud services website, where the service description is usu-ally found. Table 1 breaks down the cloud services collection and verifica-tion results. During collection, a signif-icant portion of noisy data is present. After parsing 619,474 links, the crawler found 29,189 invalid seeds from 35,601 possible seeds for cloud services (more than 80 percent). This is largely attrib-uted to the fact that we lack standards for describing and publishing cloud ser-vices. Therefore, an urgent need exists for standardization on cloud services, such as interfacing and discovery.

Note that the total number of inactive cloud services is signifi-cantly low (only 423, or roughly 0.1 percent of the total possible seeds). Search engines regularly check out-dated links and exclude them from their indexes. For those inactive cloud services, our crawler also captured the error codes according to the RFC 2616 status code definitions from the W3C (www.w3.org/Protocols/rfc2616/rfc2616-sec10.html), as Table 2 shows.

Table 1. Breakdown of cloud services collection results.

Cloud services collection Start page WSDL/WADL Ads Total

links parsed 617,285 1,552 637 619,474

Possible seeds 34,348 616 637 35,601

Inactive 366 57 0 423

Active 34,619 559 637 35,815

Invalid 28,736 453 0 29,189

Valid 5,883* 106 637 5,883

*Cloud services identified from the Web Services Description language (WSDl)/Web Application Description language (WADl) and advertisements are also included in the results.

Table 2. Error codes for inactive cloud services.

Error Code Description Percentage

101 The connection was reset 13.66

105 unable to resolve the servers DNS address 1.64

107 Secure Sockets layer protocol error 0.27

118 The operation timed out 0.27

324 The server closed the connection without sending any data

0.27

330 Content decoding failed 0.27

400 Bad request 0.82

403 Access denied 3.83

404 The requested uRl / was not found on this server

10.11

500 Server error 1.37

503 The service is unavailable 0.27

504 Page not found 0.27

1005 uRl does not exist 66.95

Total 100

IC-18-04-WSWF.indd 58 30/05/14 5:30 PM

Analysis of Web-Scale Cloud Services

july/AuGuST 2014 59

In this table, we can see that the high-est percentage (66.95 percent) goes to error code 1005 (that is, the URL doesnt exist), which means that the majority of inactive cloud services are discontinued.

Locations and LanguagesOne of our studies about cloud ser-vices, and cloud computing in general, deals with its geographical status (that is, from which part of the world cloud services are provisioned). We extracted the country domain from each URL of the collected cloud services. When the country domain wasnt present, we exploited address lookup tools such as whois (http://ipduh.com/ipv6/whois/ or www.sixxs.net/tools/whois/) to deter-mine the URLs location, which essen-tially traces back to the geographical location of the hosting datacenter and helps us determine the cloud services country information. For presentation purposes, we group countries into dif-ferent regions for a holistic view of cloud computing trends and depict the information on a world map (Figure 2). We present details about particular

countries in a specific color, according to the percentage range of the cloud services that country provisions.

From Figure 2, we note that the North American region is the big-gest provider for cloud services, with 60.45 percent. This is followed by Europe (23.27 percent). Asia provi-sions about 8.7 percent of the cloud services (about 1 percent from the Middle East), and 5.27 percent are from Australia. The remaining 2.31 percent of the cloud services are pro-visioned from other regions, includ-ing South America and Africa.

We also conducted some statistics on the languages used for the collected cloud services. For this task, we lever-aged online tools What Language Is This (http://whatlanguageisthis.com) and an open source system called Language Detection Library for Java (http://code.google.com/p/language-detection/). Figure 3a shows the sta-tistical information of the languages that are used in the cloud services. From the figure, it is clear that most cloud service providers use English (85.33 percent). This is consistent with

the fact that a large portion of cloud services are provided by countries in North America, Australia, and Europe, and most of them are English speak-ing. We can see from other languages used in cloud services such as Chi-nese, French, German, and Spanish that cloud computing is achieving broad adoption. Noticeably, 4.30 per-cent of cloud services are in an Arabic language.

Cloud Service ProviderCategorizationCloud services are widely categorized as IaaS, PaaS, or SaaS, provisioned by different cloud service providers. Determining the percentages of differ-ent kinds of service providers would be interesting. As described, after our CSCE finishes validating cloud service seeds, it categorizes the cloud services into IaaS, PaaS, or SaaS by reasoning over the relations between the con-cepts in the cloud services ontology.

Figure 3b depicts the categoriza-tion results, with providers catego-rized into six different categories: IaaS, PaaS, SaaS, IaaS+PaaS, IaaS+SaaS,

Figure 2. Cloud services locations. This gives us a holistic view of cloud computing trends worldwide.

>6%

North America60.45%

South America1.04%

Africa1.27%

Europe23.27%

Asia8.7%

Australia5.27%

0.5%6%0.1%0.5%0.05%0.1%

Web-Scale Workflow

60 www.computer.org/internet/ IEEE INTERNET COMPuTING

PaaS+SaaS, and all. Note that when a cloud service provider is categorized as IaaS+PaaS, this provider offers both IaaS and PaaS services. From the figure, we can see a fair degree of variation among providers. In particular, more than half (52.29 percent) focus on providing IaaS services, nearly one third (27.08 per-cent) on providing SaaS services, and 7.70 percent on providing PaaS services. The remaining 12.93 percent offer more than one cloud service model. Major players such as Microsoft, Amazon, and Google belong to this part.

Cloud Services and QoS Quality-of-service (QoS) attributes are critical in cloud service discovery. With

QoS information, we could rank col-lected cloud services according to con-sumers requirements, and always select the best ones for users or workflow applications. Our CSCE collected cloud services QoS data by visiting review websites that document consumers feedback. Among QoS attributes, were particularly interested in trust, given that its widely considered a key chal-lenge in cloud adoption.1,4,6,7

We analyzed 10,076 feedbacks col-lected from 6,982 users on 113 real cloud services. Figure 3c depicts the results. Cloud service consumers gave trust feedback in numerical form, with a range between 0 and 5, where 0 and 5 mean the most negative and the most

positive, respectively. From the figure, we can observe that the majority of cloud service consumers (62.26 percent) are positive (scoring 45) in trusting the cloud services they used. Only 20.06 percent of cloud service consumers were negative (scoring between 02) in trust-ing cloud services, and the rest (17.68 percent) of the feedback was neutral.

Cloud Services and SOCService-oriented computing (SOC) and Web services are one of the most impor-tant enabling technologies for cloud computing.2,5,8 Thus, we wanted to investigate SOC adoption in cloud com-puting. We conducted some preliminary studies based on the information we collected. We first investigated how much SOC description languages such as WSDL or WADL have been used for publishing cloud services. To do this, we compared the number of cloud services that have WSDL or WADL documents (for SOAP-based or RESTful Web ser-vices, respectively).

Figure 3d depicts the result. We were surprised to discover that only a very small portion of cloud services (merely 1.80 percent) were implemented using Web service interface languages. How-ever, our crawler might not detect cloud services that actually used SOC because not all WSDL documents are publicly accessible on the Internet.9 In addition, the majority of RESTful Web services provide no formal descriptions and rely on informal documentation.10 Never-theless, the low percentage still indi-cates poor adoption of SOC in cloud computing.

Advertisements and IPsWe also investigated how cloud services advertise themselves so that potential customers can find them. Search engines not only index cloud services, but some providers also advertise their services via this medium. These advertisements are usually located on the top or to the right of the returned search pages. Accordingly, our CSCE collected these advertised cloud services. Figure 3e

Figure 3. Statistical results and analysis. We looked at (a) languages used in cloud services; (b) cloud service provider categorization; (c) cloud service consumer trust feedback; (d) cloud services in the Web Service Description Language (WSDL) or Web Application Description Language (WADL); (e) cloud services advertised on search engines; and (f) cloud services IPs.

85.33%

4.30%2.80%0.95%

0.92% 0.85%4.85%

EnglishArabicChineseSpanishGermanFrenchOthers

02345

52.29%

7.70%

27.08%

2.09%6.46%

1.75%2.63%

17.68%

62.26%

20.06%

98.20%

1.80%

89.20%

10.80%

97.42%

2.58%

IaaSPaaSSaaSIaaS+PaaSIaaS+SaaSPaaS+SaaSAll

Non-WSDL & WADL

WSDL & WADL

AdvertisedNon-advertised

IPv6IPv4

(a) (b)

(c) (d)

(e) (f)

IC-18-04-WSWF.indd 60 30/05/14 5:30 PM

Analysis of Web-Scale Cloud Services

july/AuGuST 2014 61

shows that about 10.80 percent of col-lected cloud services use paid adver-tisements as a means for customers to discover them. Because advertised cloud services rely only on a short descrip-tion text to introduce themselves, user queries that normally require more information (for example, functions and QoS information) cant be answered via these advertisements.

Another interesting and important aspect worth investigating is the cloud services communication (that is, what type of IPs do cloud services use?). We used an nslookup command to deter-mine what type of IP cloud services are using (IPv4 or IPv6). We wrote a sim-ple Java program to enable automatic retrieval of such IP addresses from the collected URLs. As Figure 3f shows, most cloud services (97.42 percent) use IPv4. This makes sense because IPv4 is still the most widely deployed Internet- layer protocol.

T he most intriguing finding in our evaluation is that SOC isnt play-ing a significant role in enabling cloud computing as a technology; this is contrary to whats documented in current literature. More investigation is needed to understand why this is the case and how to enable SOC to contribute toward cloud computing so as to capitalize on previous efforts in R&D in SOC communities.

In addition, the lack of standard-ization in current cloud products and services makes cloud services discov-ery a more difficult task and a bar-rier for scalable and unified access to such services. An urgent need thus exists for standardization, especially in description languages, before we can fully embrace cloud computing. Fortunately, the research community is making some attempts at standard-ization, and has achieved some ini-tial results. For example, in August 2012, the Distributed Management Task Force (DMTF) released the Cloud Infrastructure Management Interface

(CIMI) specification, which standard-izes interactions between cloud envi-ronments to achieve interoperable cloud infrastructure management (www. dmtf.org/news/pr/2012/8/dmtf- releases-specication-simplifying-cloud-infrastructure-management).

To the best of our knowledge, ours is the first effort in discovering, col-lecting, and analyzing cloud services on a Web scale. The collected data-sets (1.06 Gbytes of metadata), which are available at http://cs.adelaide.edu.au/~cloudarmor/ds.html, will bring significant benefits to the cloud ser-vice research community.

Our ongoing research includes further investigating the relationship between SOC and cloud computing by discovering more evidence on cloud services implemented using SOC tech-nology. We also plan to extend the CSCE to perform more comprehensive QoS metrics to rank cloud services.

Acknowledgments Talal H. Noors work is supported by King

Abdullahs Postgraduate Scholarships, the Min-

istry of Higher Education: Kingdom of Saudi

Arabia. Quan Z. Shengs work is partially sup-

ported by Australian Research Council Discov-

ery grant DP130104614 and DP140100104. We

thank Jeriel Law and Abdullah Alfazi for their

participation in data collection.

References 1. M. Armbrust et al., A View of Cloud Com-

puting, Comm. ACM, vol. 53, no. 4, 2010,

pp. 5058.

2. Y. Wei and M.B. Blake, Service-Oriented

Computing and Cloud Computing: Chal-

lenges and Opportunities, IEEE Internet

Computing, vol. 14, no. 6, 2010, pp. 7275.

3. K. Ren et al., Security Challenges for the

Public Cloud, IEEE Internet Computing,

vol. 16, no. 1, 2012, pp. 6973.

4. T.H. Noor et al., Trust Management of

Services in Cloud Environments: Obstacles

and Solutions, ACM Computing Surveys,

vol. 46, no. 1, 2013, article no. 12.

5. B. Satzger et al., Winds of Change: From Ven-

dor Lock-In to the Meta Cloud, IEEE Internet

Computing, vol. 17, no. 1, 2013, pp. 6973.

6. T.H. Noor and Q.Z. Sheng, Credibility-Based

Trust Management for Services in Cloud

Environments, Proc. 9th Intl Conf. Service-

Oriented Computing, 2011, pp. 328343.

7. K. Hwang and D. Li, Trusted Cloud Com-

puting with Secure Resources and Data Col-

oring, IEEE Internet Computing, vol. 14,

no. 5, 2010, pp. 1422.

8. T.H. Noor and Q.Z. Sheng, Trust as a Ser-

vice: A Framework for Trust Management

in Cloud Environments, Proc. 12th Intl

Conf. Web Information System Eng., 2011,

pp. 314321.

9. E. Al-Masri and Q. Mahmoud, Investigat-

ing Web Services on the World Wide Web,

Proc. 17th Intl Conf. World Wide Web,

2008, pp. 795804.

10. D. Renzel et al., Todays Top RESTful Ser-

vices and Why They Are Not RESTful, Proc.

Web Information Systems Engineering, LNCS

7651, 2012, pp. 354367.

Talal H. Noor is an assistant professor in the

Department of Computer Science at Taibah

University, Yanbu. He has a PhD in com-

puter science from the University of Ade-

laide. Contact him at tnoor@taibahu.edu.sa.

Quan Z. Sheng is an associate professor and

head of the Advanced Web Technologies

Research Group in the School of Computer

Science at the University of Adelaide. He

has a PhD in computer science from the

University of New South Wales. Contact

him at qsheng@cs.adelaide.edu.au.

Anne H.H. Ngu is a full professor in the Depart-

ment of Computer Science at Texas State

University. She has a PhD in computer sci-

ence from the University of Western Aus-

tralia. Contact her at angu@txstate.edu.

Schahram Dustdar is a full professor of computer

science and head of the Distributed Systems

Group, Institute of Information Systems, at

the Vienna University of Technology. He is

an ACM Distinguished Scientist and IBM

Faculty Award recipient. Contact him at

dustdar@dsg.tuwien.ac.at.

Selected CS articles and columns are also available for free at http://

ComputingNow.computer.org.

IC-18-04-WSWF.indd 61 30/05/14 5:30 PM